This invention relates in general to microelectronic circuits, and in particular, an interconnect structure and a method and apparatus for forming connections to microelectronic devices and between such devices.
The geometry of integrated circuits continues to shrink. At present, the distance between adjoining features in microelectronic devices is less than one micron. As CMOS gate lengths shrink, channel dopings must increase in order to maintain reasonable punchthrough voltages. With such increasing channel dopings, gate oxides are decreased in thickness, especially for faster gate switching and to maintain an approximately constant body effect. It is common in such CMOS technology to use polysilicon electrodes, usually N doped, with phosphorous, for forming the gate electrode. The work function of N+ polysilicon is about 4.05 electron volts which produces an accumulated surface charge on N type silicon resulting in larger than desired PMOS threshold. Another unwanted effect is the depleted surface on P type silicon which produces lower than desired NMOS thresholds. In order to increase the PMOS thresholds, a shallow boron implant is required. However, such counter doping produces a buried-channel type device that has poor turnoff subthreshold characteristics and is especially subject to short channel effects for submicron gate lengths.
In order to solve the above problems, in the past, others have used higher work function gate electrodes, for example, molybdenum (about 4.7 eV) or have proposed using P+ poly (about 5.25 eV) for PMOS devices. Another solution has been to lengthen the PMOS gate. However, lengthening the gate reduces the AC performance of the PMOS device which is already lower than its companion NMOS device.
It is known in CMOS technology to salicide polysilicon gates and sources and drains of CMOS devices. Such salicidation (simultaneously siliciding gates, sources and drains) reduces the resistance of these electrodes. With reduced resistance, devices can be made faster since the charging of the junctions and parasitic capacitance is done through smaller resistances.
The silicide of choice in many applications is titanium silicide. After the silicide process, unreacted titanium is conveniently reacted with a nitrogen-containing gas to form titanium nitride. Thus a thin layer of conductive titanium nitride (TiN) layer is suitable for forming local interconnects without a further deposition step. However, using TiN as a local interconnect has problems. First of all, its sheet resistance is marginal (4-40 ohms per square). For another, its formation process is sensitive to moisture and oxygen in the reaction chamber both of which are difficult to control. Also, resist adhesion is sometimes a problem, i.e., photoresist may not adhere to a TiN layer. It is often difficult to etch a TiN layer because of the low selectivity between a TiN layer and its underlying silicide. Underetch results in stringers or filaments which cause shorts; overetch erodes the silicide from the sources and drains causing high source-drain resistance.
Thus, there exists a problem in the current state of the art for forming a local interconnect which avoids the inherent problems associated with titanium nitride local interconnects and the penalties of using doped P implants that form buried channel type devices.